Gain-Scheduled Steering Control For Autonomous Vehicles

This study presents a linear parameter varying (LPV) approach for the lateral control of autonomous vehicles, in order to take into account the whole operating domain of longitudinal speeds, as well as the variation of the look-ahead distance. Combining a dynamical vehicle model with look-ahead dynamics, together with an identified actuator model including an input delay, the closed-loop performances can be achieved and the tracking capabilities can be improved for every speed. This is obtained in particular through ad hoc representation of the look-ahead time as a parameter-dependent function. An H-infinity LPV control problem is formulated considering parameter-dependent weighting functions, allowing the control adaptation for all speeds. The synthesis is performed using the gridding approach, in order to account for varying parameter rate. The proposed steering control system has been implemented on a real electric Renault Zoe car. The performances are therefore assessed experimentally on a real test track with a varying longitudinal speed profile, and compared with a classical LPV polytopic controller, which proves the advanced lane-tracking capabilities of the proposed methodology.